Fixes#1630 where --redo-ocr would shift OCR text vertically on PDFs
with non-zero mediabox origins (e.g., [0, 100, width, height+100]).
The bug occurred in _graft_fpdf2_text_layer where the Form XObject BBox
was set to the text layer's mediabox [0, 0, w, h] instead of the base
page's mediabox [0, 100, w, h+100]. This caused a coordinate mismatch
between the BBox and the transformation matrix, resulting in text being
positioned incorrectly.
The fix changes line 450 in _graft.py to use base_mediabox instead of
mediabox, making the fpdf2 renderer consistent with the sandwich renderer
which already used base_mediabox correctly.
This issue commonly affected:
- JSTOR PDFs (generated by iText with cropping)
- Cropped PDFs from various tools
- PDFs with non-standard coordinate systems
Added regression test that creates a PDF with offset mediabox origin
and verifies --redo-ocr preserves coordinates correctly.
strip_invisible_text resets the text render mode on each `BT` (begin text) command. However the text state is not actually reset for each text element, only for each page.
The pdf reference says:
> The text state operators can appear outside text objects, and the values they set
> are retained across text objects in a single content stream. Like other graphics
> state parameters, these parameters are initialized to their default values at the
> beginning of each page.
>
> -- https://opensource.adobe.com/dc-acrobat-sdk-docs/pdfstandards/pdfreference1.7old.pdf#page=397
With the current implementation, a text object is only deleted if it contains a `3 Tr` command (setting the text rendering mode to invalid). However the rendering mode may be set once and then not changed for multiple text objects or set outside of a text object.
In that case only the first text object (which contains the `3 Tr`-command) is removed. This not only leaves the other text objects in the pdf, but also makes them visible, since the text object that contained the `3 Tr`-command is removed.
This PR updates `strip_invisible_text` to not reset the rendering mode for each object and to keep track of the rendering mode when the graphic state is pushed/popped.
James R. Barlow
Kara Engelhardt
James R. Barlow